Predictors of changes in preoperative tumor stage between dynamic computed tomography and gadoxetate disodium- enhanced magnetic resonance imaging for hepatocellular carcinoma

Background/Purpose: Gadoxetate disodium-enhanced magnetic resonance imaging (EOB-MRI) has a higher diagnostic accuracy for hepatocellular carcinoma (HCC) than computed tomogra-phy (CT). However, indications for performing EOB-MRI after dynamic CT are not well defined. Therefore, we investigated the clinical factors associated with changes in the preoperative tu-mor stage between dynamic CT and EOB-MRI.Methods: A prospective cohort was conducted from January 2014 to December 2017. 156 adult patients with clinical suspicion of HCC before liver resection were enrolled and we retrospec-tively reviewed the images. The tumor staging was evaluated by dynamic CT and then EOB-MRI subsequently according to the TNM staging system. The changes in tumor stage between two modalities were identified, and the associated clinical factors were analyzed.Results: A total of 99 patients were analyzed after excluding 57 patients. 20 patients (20.2%) had changes in tumor stage between dynamic CT and EOB-MRI. The change occurred only in early stage (T1 and T2 lesions) based on dynamic CT initially. Furthermore, in univariate and multivariate analyses, albumin-bilirubin (ALBI) grade 2 and log alpha-fetoprotein (AFP) levels were associated with changes in tumor staging by EOB-MRI than those without (50% vs. 9.9%, p < 0.001 and 2.04 +/- 1.35 vs. 1.40 +/- 1.16, p Z 0.038, respectively). Patients with changes in tumor stage also exhibited higher 1-year recurrence rate and shorter recurrence-free survival.Conclusion: Changes in preoperative tumor stage between dynamic CT and EOB-MRI were asso-ciated with CT-defined early stage, ALBI grades, higher log AFP levels, and early recurrence.Copyright 2022, Formosan Medical Association. Published by Elsevier Taiwan LLC. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).